469916 - When starting Firefox: "lcms: Error #12288; Read from memory error" / "Corrupted memory profile"

Status: RESOLVED FIXED

(Core :: Graphics: Color Management, defect, P4)

Description

[Daniel Holbert [:dholbert]]

I just noticed these scary-looking messages floating by, when I start Firefox using my mozilla-central debug build (up-to-date as of this morning -- updated to changeset fd035640d161).

lcms: Error #12288; Read from memory error. Got 0 bytes, block should be of 128 bytes
lcms: Error #12288; Corrupted memory profile

This happens both when using a completely fresh profile (running the first time, with all config options set to their defaults) and when re-using an existing profile.

I'm running Ubuntu 8.10, with a dual-monitor setup, if that matters.

Comment 1

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

I'm wondering if you maybe have a 0-byte profile atom in X, and we're blindly grabbing that and not checking its size?  Check.. um, I have no idea how to list atoms any more.  xlsatoms? xwininfo -root?  Look for an atom named _ICC_PROFILE on the root window.

Comment 2

[Daniel Holbert [:dholbert]]

Here's what I get:
  [dholbert@orange:~]$ xlsatoms -name _ICC_PROFILE
  628	_ICC_PROFILE

What next? :)  

FWIW, xwininfo -root doesn't mention either "icc" or "profile" at all.  (That is, "xwininfo -root | grep -i icc" and "xwininfo -root | grep -i profile" both yield nothing.)

Comment 3

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Hm, I think xprop now?  xprop -root _ICC_PROFILE ?

Comment 4

[Daniel Holbert [:dholbert]]

Here's the result:
  [dholbert@orange:~]$ xprop -root _ICC_PROFILE 
  _ICC_PROFILE:  not found.

and here's me just testing if that syntax works, with another atom:
  [dholbert@orange:~]$ xprop -root _NET_NUMBER_OF_DESKTOPS
  _NET_NUMBER_OF_DESKTOPS(CARDINAL) = 4

Also, if I grep "xprop -root" for "icc" or "profile", I get nothing.

Comment 5

[Bobby Holley (:bholley)]

Well, for starters the error message is actually incorrect. This was fixed in a patch that I r- ed, so it never actually got landed. See https://bugzilla.mozilla.org/attachment.cgi?id=347178&action=diff, bug 458164. Anything we do for this bug should certainly fix that message.

The reading is going on here:
http://mxr.mozilla.org/mozilla-central/source/gfx/thebes/src/gfxPlatformGtk.cpp#551
It seems pretty clear that we're doing no checking for 0-sized profiles. Whether checking for appropriately-sized profiles is thebes' job or lcms' job is an interesting question. Right now LCMS seems to be doing the right thing, though perhaps not in the nicest manner. Thoughts?

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

No reason why we shouldn't check in thebes as well.  I can cook up that patch if you're busy, let me know.

Comment 7

[Bobby Holley (:bholley)]

vlad - quarter's starting up, so that would be great - thanks

Comment 8

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: check for 0-sized return sfrom X

Ok, here's a simple patch that just ignores 0-sized profiles in X atoms.

Comment 9

[Bobby Holley (:bholley)]

Comment on attachment 355532 [details] [diff] [review]
check for 0-sized return sfrom X

looks good - thanks

Comment 10

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

23417[tip]   4f196841ba1a   2009-01-07 11:50 -0800   vladimir
  b=469916, lcms error when trying to open 0-sized profile; r=bholley